Building knowledge from address information for intelligent geocoding
Intelligent geocoding is the process of applying knowledge and information to geocoding. This paper describes a research project that is utilising intelligent processes and algorithms to link street addresses and other geographic descriptions to spatial location. User and application contexts requiring intelligent geocoding are wide-ranging and include emergency response, routing and navigation, demographic marketing, customer location identification and analysis, health services management, epidemiology, etc.
Geocoding is the process of identifying “geographic location” from textual data such as an address or feature name. Geocoding queries can be very complex to execute and depend very much on the address information (supplied by users or applications) available, the quality of the address request, the reliability of address reference information, the purpose of the user, the application context of the query, and the geographic context of the resulting spatial location.
For most of the common and standard addresses used, this process of geocoding can be relatively simple, but where address information is missing, incomplete, ambiguous, incorrect, uses aliases, etc. the process may become extremely difficult. The address information that exists across several data sources (ie. national address reference files, state-based land parcel data, user’s head, etc.) needs to be integrated into a knowledge base and associated with other information related to address locations, to derive meaningful information and knowledge and resolve spatial locations.
This research focuses on a solution for intelligent geocoding that utilises an agent-based system supported by a knowledge base. Multiple agents collaborate in the geocoding process and access knowledge from the knowledge base, together with information from external sources, to resolve the spatial location of an address. This paper details the methodology used and describes how the knowledge base is used to support the decision-making tasks of the agents.